APPARATUS AND METHOD FOR ENCODING AND DECODING A PICTURE USING

PICTURE BOUNDARY HANDLING

GENERAL DESCRIPTION

The present invention is concerned with picture and video coding. Embodiments of the present invention are concerned with a particular way of partitioning input picture and video data of a specific size into smaller entities. In particular, embodiments of the present invention are concerned with picture boundary handling in recursive picture signal partitioning.

BACKGROUND OF THE INVENTION

In modern video coding applications, the input video signal of a specific size is partitioned into smaller chunks [1], This partitioning consists of multiple structures with specific information and description associated with each level. In the state of the art video codec H.265/HEVC [1], the most important subdivision is the subdivision into macroblocks. Those macroblocks, or coding tree units (CTUs), are quadratic structures of a predefined size (e.g. 32x32 or 64x64 luma samples) spanning a fixed grid over the coded picture. All other partitioning levels are defined in terms of this rigid subdivision. E.g., the more coarse high level partitioning into slices and tiles is defined in terms of the included CTUs.

In H.265/HEVC [1], starting at each CTU, a quad-tree structure is signaled providing means of recursively partitioning a rigid CTU into flexible sub-structures. At each level, in a defined parameter scope, it is inferred or signaled a block that should be split into four sub-blocks or coded at the specified level. This process is repeated recursively until no further split is possible as defined by current high-level parameters or a flag is read indicating that no further split should be performed at the current level. In this case, signaling of a coding unit and its sub-structures will be read.

The signaling of a quad-tree split can sometimes be omitted and inferred as either true or false. Most notably, due the rigid nature, CTUs are not necessarily aligned with video pic-ture boundaries. In this case, if a CTU or a sub-block in the quad-tree partitioning structure is not fully contained in the picture boundaries, a split flag will be inferred as positive without explicit signaling. If, after the split, a sub-block lies completely outside of the video picture boundaries, no further signaling for this block will be read.

In the development of a future video standard with capabilities beyond H.265/HEVC [1], the quad-tree recursive splitting structure is extended by different split configurations [2], If a block is not split into four quadratic sub-parts, a binary split is signaled indicating that this block should be split into two rectangular blocks using a split ratio of ½. The signaling also includes the information if the split should be applied horizontally or vertically. Such rectangular blocks can be further recursively split into smaller quadratic or rectangular blocks using the binary tree split syntax. This enhanced split-tree is called QTBT [2],

QTBT handles picture boundaries using the implicit quad-split, exactly as in the

H.265/HEVC standard [1] In QTBT, if a block exceeds the picture boundaries, a quad-split will be inferred and no signaling will be read.

In [5], a novel partitioning method is described, Generalized Binary Splitting (GBS), which might not contain a quad-split but might also be used together with a quad split. In this method the binary split can be signaled to perform the split with a split ratio other than ½ (in QTBT, the binary split performs the split in the middle of the current block). The splits might also be signaled using a syntax relative to the previous split (e.g. perpendicular and parallel rather than horizontal and vertical).

However, the conventional partitioning methods are rather rigid in the handling of picture boundaries. As explained above, conventional partitioning methods may use an implicit quad split at picture boundaries. That is, the prior art uses a fully implicit split derivation at a picture boundary. Such implicit signaling may enable a good signaling efficiency. However, this may only provide for a low flexibility.

Thus, it is an object of the present invention to enhance existing partitioning methods to be more flexible in picture boundary handling and, at the same time, to reduce the bit budget for signaling the picture boundary handling.

The present invention offers a solution by providing an apparatus for encoding a picture according to claim 1 , an apparatus for decoding a picture according to claim 10, a method for encoding a picture according to claim 19, a method for decoding a picture according to claim 20, and a computer program according to claim 21.

A first aspect concerns an apparatus for encoding a picture. The apparatus is configured to partition the picture into leaf blocks using a recursive multi-tree partitioning. That is, the apparatus is configured to partition the picture into smaller chunks using said recursive multi-tree partitioning. Said smaller chunks may be blocks of a certain size. By stepping through the multi-tree, the partitioning of the picture may start from a tree-root block at a first level and it may end up at a leaf block, which is the last block of the multi-tree and thus the smallest entity of the partitioning. Between the tree-root block and the partitioned leaf block, the apparatus may step-by-step go from one partitioning tree level to one or more subsequent partitioning tree levels, wherein at each tree level the current block at the particular tree level is partitioned into two or more smaller blocks. For example in HEVC, the tree-root blocks may be so-called macroblocks or Coding Tree Units (CTUs), and the leaf blocks may be so-called subblocks or Coding Unit (CUs). Accordingly, a CTU may be partitioned into one or more CUs, wherein a leaf block or any block at a tree level between the tree-root block and the leaf block may be referred to as a subblock. Furthermore, in the present disclosure, a tree-root block, or a leaf block, or any block at a tree level between the tree-root block and the leaf block may be referred to as a predetermined block. Accordingly, a predetermined block may correspond to a predetermined tree level of the multi-tree partitioning. The apparatus may exploit a block-based coding scheme, i.e. the apparatus is configured to block-based encode the picture into a data stream by using the partitioning of the picture into the leaf blocks. The partitioning may also be referred to as splitting. The partitioning of the tree-root blocks into the smaller leaf blocks may use a certain splitting scheme for splitting blocks into smaller subblocks. These splitting schemes may also be referred to as split modes which may vary at each tree level. Furthermore, the predetermined blocks may comprise a predetermined size. When arranging the predetermined blocks as a grid over the picture, it may happen that some of the predetermined blocks, due to its size, may extend over a picture boundary. For example, a first portion of a predetermined block may be located inside the picture while a second portion of said predetermined block may be located outside the picture. Accordingly, the picture boundary may cross the predetermined block. Generally, the picture boundary may cross predetermined blocks at different positions. For example, the picture boundary may cross a predetermined block horizontally, or vertically, or both wherein the picture corner will be contained in the predetermined block. For handling these situations, the present invention provides the following solution. While partitioning the picture into the leaf blocks, the apparatus is configured to partition a predetermined block that corresponds to a predetermined tree level of the multi-tree partitioning and which extends beyond a boundary of the picture, by using a reduced set of split modes compared to fully implicit split derivation as used by the prior art. According to the invention, the apparatus is configured to reduce an available set of split modes for splitting the predetermined block depending on the above mentioned position at which the picture boundary crosses said predetermined block. Thus, the apparatus obtains a reduced set of one or more split modes. The reduced set may comprise a cardinality which indicates the number of split modes

being available in the reduced set of split modes. Accordingly, if a cardinality of the reduced set is one, i.e. if the reduced set comprises only one split mode, the apparatus is configured to apply this split mode of the reduced set for splitting the predetermined block. If a cardinality of the reduced set is greater than one, i.e. if the reduced set comprises two or more different split modes, the apparatus is configured to select one of these split modes of the reduced set and to apply the selected one of the split modes for splitting the predetermined block, wherein the apparatus signals its respective selection in the data stream. In other words, the inventive apparatus may pre-select a reduced set of split modes for splitting a current predetermined block. Said pre-selection may depend on the current position of the block relative to the picture boundary. In result, the inventive apparatus may only have to choose a suitable split mode from a pre-selected reduced set of split modes. This reduces the bit budget for signaling the selected split mode in the data stream because, no information about splitting itself needs to be transferred, and if the reduced split set comprises more than one split mode, only the remaining uncertainty may have to be signaled, e.g. by one bin.

A second aspect concerns an apparatus for decoding a picture. The apparatus is config-ured to partition the picture into leaf blocks using recursive multi-tree partitioning. The apparatus is further configured to block-based decode the picture from a data stream using the partitioning of the picture into leaf blocks. The apparatus is further configured to, in partitioning the picture into the leaf blocks, for a predetermined block that corresponds to a predetermined tree level of the multi-tree partitioning and that extends beyond a boundary of the picture, reduce an available set of split modes for splitting the predetermined block depending on a position at which the boundary crosses the predetermined block to obtain a reduced set of one or more split modes. If a cardinality of the reduced set is one, the apparatus is configured to apply the split mode of the reduced set for splitting the predetermined block, and if a cardinality of the reduced set is greater than one, the apparatus is configured to select one of the split modes of the reduced set and to apply the selected one of the split modes for splitting the predetermined block according to a signalization in the data stream. As to the advantages of said apparatus for decoding a picture it is referred to the passage above describing the advantages of the apparatus for encoding a picture.

A third aspect concerns a method for encoding a picture, the method comprising a step of partitioning the picture into leaf blocks using recursive multi-tree partitioning. The method further comprises a step of block-based encoding the picture into a data stream using the partitioning of the picture into the leaf blocks. The method further comprises the steps of, in partitioning the picture into the leaf blocks, for a predetermined block that corresponds to a predetermined tree level of the multi-tree partitioning and which extends beyond a boundary of the picture, reducing an available set of split modes for splitting the predetermined block depending on a position at which the boundary crosses the predetermined block to obtain a reduced set of one or more split modes. If a cardinality of the reduced set is one, the method comprises a step of applying the split mode of the reduced set for splitting the predetermined block, and if a cardinality of the reduced set is greater than one, the method comprising a step of selecting one of the split modes of the reduced set and applying the selected one of the split modes for splitting the predetermined block and signaling the selection in the data stream. As to the advantages of said method for encoding a picture it is referred to the passage above describing the advantages of the apparatus for encoding a picture.

A fourth aspect concerns a method for decoding a picture, the method comprising a step of partitioning the picture into leaf blocks using recursive multi-tree partitioning. The method further comprises a step of block-based decoding the picture from a data stream using the partitioning of the picture into leaf blocks. The method further comprises the steps of, in partitioning the picture into the leaf blocks, for a predetermined block that corresponds to a predetermined tree level of the multi-tree partitioning and which extends beyond a boundary of the picture, reducing an available set of split modes for splitting the predetermined block depending on a position at which the boundary crosses the predetermined block to obtain a reduced set of one or more split modes. If a cardinality of the reduced set is one, the method comprises a further step of applying the split mode of the reduced set for splitting the predetermined block, and if a cardinality of the reduced set is greater than one, the method comprising the further step of selecting one of the split modes of the reduced set and applying the selected one of the split modes for splitting the predetermined block according to a signalization in the data stream. As to the advantages of said method for decoding a picture it is referred to the passage above describing the advantages of the apparatus for encoding a picture.

According to a fifth aspect, computer programs are provided, wherein each of the computer programs is configured to implement the above-described method when being exe-cuted on a computer or signal processor, so that the above-described method is implemented by one of the computer programs.

Advantageous aspects of the present invention are the subject of the independent claims, Preferred embodiments of the present application, which are defined in the dependent claims, will be exemplarily described below with respect to the figures, wherein

Fig, 1 shows a block diagram of an apparatus for predictively coding a picture as an example for an encoder where an intra prediction concept according to embodiments of the present application could be implemented,

Fig. 2 shows a block diagram of an apparatus for predictively decoding a picture, which fits to the apparatus of Fig. 1 , as an example for decoder where an intra prediction concept according to embodiments of the present application could be implemented,

Fig. 3 shows a schematic diagram illustrating an example for a relationship be- tween the prediction residual signal, the prediction signal and the recon- structed signal so as to illustrate possibilities of setting subdivisions for coding mode selection, transform selection and transform performance, respectively,

Claims: 1. Apparatus (10) for encoding a picture (12), configured to
partition the picture (12) into leaf blocks using recursive multi-tree partitioning,
block-based encode the picture (12) into a data stream (14) using the partition-ing of the picture (12) into the leaf blocks,
wherein the apparatus (10) is configured to, in partitioning the picture (12) into the leaf blocks,
for a predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) that corresponds to a predetermined tree level of the multi-tree partitioning and which extends beyond a boundary of the picture (12), re-duce an available set of split modes for splitting the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) depending on a posi-tion at which the boundary of the picture (12) crosses the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) in order to ob-tain a reduced set of one or more split modes,
wherein, the apparatus (10) is configured to, for reducing the available set of split modes to the reduced set of one or more split modes, select from the available set of split modes at least one of
a quad split, and
a bi-split that splits the predetermined block (801a; 802d) into two sub-blocks (802a, 802b; 803b, 803d) along a split line (811; 813) that is parallel to the picture boundary’s position in the predetermined block (801a; 802d) and which split line (811; 813) is at least one of collocated to the picture boundary and centered inside the predetermined block (801a; 802d).

2. The apparatus (10) of claim 1, wherein the apparatus (10) is configured to, for reducing the available set of split modes to the reduced set of one or more split modes, select from the available set of split modes at least one of
a first split mode which does not impose any restriction onto one or more sub-sequent splits for subsequent tree levels succeeding the predetermined tree level, and
a second split mode that splits the predetermined block (801a) into two sub-blocks (802a, 802b) along a split line (811) that is parallel to the picture boundary’s po-sition in the predetermined block (801a) and which split line (811) is at least one of collocated to the boundary of the picture (12) and centered inside the predetermined block (801a).

3. The apparatus (10) of claim 1, wherein the apparatus (10) is configured to, for reducing the available set of split modes to the reduced set of one or more split modes, select from the available set of split modes
one predetermined split mode that splits the predetermined block (801a; 802d) into two subblocks (802a, 802b; 803b, 803d) along a first split line (811; 813) that is parallel to the picture boundary’s position in the predetermined block (801a; 802d) and which split line (811; 813) is at least one of collocated to the boundary of the picture (12) and centered inside the predetermined block (801a).

4. The apparatus (10) of claim 3, wherein the predetermined split mode is either a vertical bi-split or a horizontal bi-split.

5. The apparatus (10) of one of claims 1 to 4, wherein the apparatus (10) is configured to determine the available set of split modes for the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c)
independent from a location of the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) relative to the picture boundary, and/or
dependent on a sequence of split modes of preceding tree levels from which the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) emerges.

6. The apparatus (10) of any one of claims 1 to 5, wherein the apparatus (10) is config-ured to determine the available set of split modes for the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c)
to be equal to a primitive set of split modes including a quad split, at least one horizontal bi-split and at least one vertical bi-split, if all split modes of preceding tree levels from which the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) emerges are outside a restrictive set of bi-splits, and
to be equal to the primitive set of split modes less the quad split, if a split mode of a preceding tree level from which the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) emerges is one of the restrictive set of bi-splits.

7. The apparatus (10) of claim 1, wherein, if a bi-split from the reduced set is applied and the number of consecutive bi-splits is restricted to a predetermined maximum number of consecutive bi-splits, the apparatus (10) is configured to not count a consecutive bi-split if said consecutive bi-split is applied over a picture boundary.

8. The apparatus (10) of claim 6, wherein the apparatus (10) is configured to, for reducing the available set of split modes to the reduced set of one or more split modes, select from the available set of split modes the quad split only if the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) extends beyond a cor-ner of the picture 12.

9. Apparatus (20) for decoding a picture (12), configured to
partition the picture (12) into leaf blocks using recursive multi-tree partitioning,
block-based decode the picture (12) from a data stream (14) using the partition-ing of the picture (12) into leaf blocks,
wherein the apparatus (20) is configured to, in partitioning the picture (12) into the leaf blocks,
for a predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) that corresponds to a predetermined tree level of the multi-tree partitioning and which extends beyond a boundary of the picture (12), re-duce an available set of split modes for splitting the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) depending on a posi-tion at which the picture boundary crosses the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) to obtain a reduced set of one or more split modes, wherein the apparatus (20) is configured to, for re-ducing the available set of split modes to the reduced set of one or more split modes, select from the available set of split modes at least one of
a quad split, and
a bi-split that splits the predetermined block (801a) into two subblocks (802a, 802b) along a split line (811) that is parallel to the picture boundary’s position in the predetermined block (801a) and which split line (811) is at least one of collocated to the picture boundary and centered inside the predeter-mined block (801a).

10. The apparatus (20) of claim 9, wherein the apparatus (20) is configured to, for reducing the available set of split modes to the reduced set of one or more split modes, select from the available set of split modes at least one of
a first split mode which does not impose any restriction onto one or more sub-sequent splits for subsequent tree levels succeeding the predetermined tree level, and
a second split mode that splits the predetermined block (801a) into two sub-blocks (802a, 802b) along a split line (811) that is parallel to the picture boundary’s po-sition in the predetermined block (801a) and which split line (811) is at least one of collocated to the picture boundary and centered inside the predetermined block (801a).

11. The apparatus (20) of claim 9, wherein the apparatus (20) is configured to, for reducing the available set of split modes to the reduced set of one or more split modes, select from the available set of split modes
one predetermined split mode that splits the predetermined block (801a; 802d) into two subblocks (802a, 802b; 803b, 803d) along a first split line (811; 813) that is parallel to the picture boundary’s position in the predetermined block (801a; 802d) and which split line (811; 813) is at least one of collocated to the boundary of the picture (12) and centered inside the predetermined block (801a; 802d).

12. The apparatus (10) of claim 11, wherein the predetermined split mode is either a verti-cal bi-split or a horizontal bi-split.

13. The apparatus (20) of one of claims 9 to 12, wherein the apparatus (20) is configured to determine the available set of split modes for the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c)
independent from a location of the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) relative to the picture boundary, and/or
dependent on a sequence of split modes of preceding tree levels from which the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) emerges.

14. The apparatus (20) of any one of claims 9 to 13, wherein the apparatus (20) is config-ured to determine the available set of split modes for the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c)
to be equal to a primitive set of split modes including a quad split, at least one horizontal bi-split and at least one vertical bi-split, if all split modes of preceding tree levels from which the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) emerges are outside a restrictive set of bi-splits, and
to be equal to the primitive set of split modes less the quad split, if a split mode of a preceding tree level from which the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) emerges is one of the restrictive set of bi-splits.

15. The apparatus (20) of claim 9, wherein, if a bi-split from the reduced set is applied and the number of consecutive bi-splits is restricted to a predetermined maximum number of consecutive bi-splits, the apparatus (20) is configured to not count a consecutive bi-split if said consecutive bi-split is applied over a picture boundary.

16. The apparatus (20) of claim 9, wherein the apparatus (20) is configured to, for reducing the available set of split modes to the reduced set of one or more split modes, select from the available set of split modes the quad split only if the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) extends beyond a cor-ner of the picture (12).

17. A method for encoding a picture (12), the method comprising the steps of
partitioning the picture (12) into leaf blocks using recursive multi-tree partition-ing,
block-based encoding the picture (12) into a data stream (14) using the parti-tioning of the picture (12) into the leaf blocks,
wherein the method further comprises the steps of, in partitioning the picture (12) into the leaf blocks,
for a predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) that corresponds to a predetermined tree level of the multi-tree partitioning and which extends beyond a boundary of the picture (12), re-ducing an available set of split modes for splitting the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) depending on a position at which the picture boundary crosses the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) in order to obtain a re-duced set of one or more split modes, wherein reducing the available set of split modes to the reduced set of one or more split modes comprises selecting from the available set of split modes at least one of
a quad split, and
a bi-split that splits the predetermined block (801a) into two subblocks (802a, 802b) along a split line (811) that is parallel to the picture boundary’s position in the predetermined block (801a) and which split line (811) is at least one of collocated to the picture boundary and centered inside the predeter-mined block (801a).

18. A method for decoding a picture (12), the method comprising the steps of
partitioning the picture (12) into leaf blocks using recursive multi-tree partition-ing,
block-based decoding the picture (12) from a data stream (14) using the parti-tioning of the picture (12) into leaf blocks,
wherein the method further comprises the steps of, in partitioning the picture (12) into the leaf blocks,
for a predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c), which corresponds to a predetermined tree level of the multi-tree parti-tioning and which extends beyond a boundary of the picture (12), reducing an availa-ble set of split modes for splitting the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) depending on a position at which the picture boundary crosses the predetermined block (801a, 801b, 801c, 801d, 802a, 802b, 802d, 803a, 803b, 803c) in order to obtain a reduced set of one or more split modes, wherein reducing the available set of split modes to the reduced set of one or more split modes comprises selecting from the available set of split modes at least one of
a quad split, and
a bi-split that splits the predetermined block (801a) into two subblocks (802a, 802b) along a split line (811) that is parallel to the picture boundary’s position in the predetermined block (801a) and which split line (811) is at least one of collocated to the picture boundary and centered inside the predetermined block (801a).

19. A data stream obtained by the method of claims 17 or 18.